I'm sure that the math relation is know. Don't try to mystify the subject - thats typically high-end mumbo-jumby logic to sell things that they don't even understand themselfs. Jitter creates either noise or distorsion depending on if correlated with the music signal or not. This is known. Should we trust your impressions and hearing ability? Why? You obviously "sell" something, not to a big profit, but you still invest your creditability and person. So something is at steak.
"We respect those who runs delta-sigma DAC" - No you don't as it is not respectable that you tell people with S/D DAC that its best used as a sausage grill - you did this several times. Bad sport. Maybe an excuse is in place as you seem to have changed you mind regarding this?
Whats the use of a good clock if the bits don't end up right? It's really a X-Y axis game - you will not aim correctly with one axis super precise and the other not so much...
You will never be able to prove anything here. Nor here or there. You *belive* in your chosen path (time) - stick to it and be confident that it works.
Some of us will also believe and - invest. Then comes the feedback - from a verity of system environment and taste. It's all a mess really :-D
Why are you so defensive when you have such a fine product? I cant in my mind understand why you keep arguing instead of just being cool and deliver thE products - your waiting list is now long. Keep cool man.
I'ts fun - and thats what matters.
//
"We respect those who runs delta-sigma DAC" - No you don't as it is not respectable that you tell people with S/D DAC that its best used as a sausage grill - you did this several times. Bad sport. Maybe an excuse is in place as you seem to have changed you mind regarding this?
Whats the use of a good clock if the bits don't end up right? It's really a X-Y axis game - you will not aim correctly with one axis super precise and the other not so much...
You will never be able to prove anything here. Nor here or there. You *belive* in your chosen path (time) - stick to it and be confident that it works.
Some of us will also believe and - invest. Then comes the feedback - from a verity of system environment and taste. It's all a mess really :-D
Why are you so defensive when you have such a fine product? I cant in my mind understand why you keep arguing instead of just being cool and deliver thE products - your waiting list is now long. Keep cool man.
I'ts fun - and thats what matters.
//
Maybe I have not explained myself well, I have nothing to sell, I have no commercial interests like other members of this forum, at most I can hope to share the development costs.
I don't want you to trust my impressions, I urge you to try yourself and then decide, or you can stay with the Crystek and live happy.
And I have never changed my opinion about delta-sigma DACs, I don't like how they sound, but of course everyone is free to use them, while I'm free to use them as a microwave oven.
Ultimately I'm not defensive at all, I have tried to explain that we have developed these devices following our approach, and so we cannot design something thinking to a delta-sigma DAC because we will never use it.
This is exactly what those with commercial interests would do but it is not our case.
I don't want you to trust my impressions, I urge you to try yourself and then decide, or you can stay with the Crystek and live happy.
And I have never changed my opinion about delta-sigma DACs, I don't like how they sound, but of course everyone is free to use them, while I'm free to use them as a microwave oven.
Ultimately I'm not defensive at all, I have tried to explain that we have developed these devices following our approach, and so we cannot design something thinking to a delta-sigma DAC because we will never use it.
This is exactly what those with commercial interests would do but it is not our case.
You have not understood, I don't know much about the theory of delta-sigma DAC, but I well know how they sound, bad IMHO.
Or do you think one has to know perfectly the theory to understand how they sound?
But honestly I don't understand, do you want to convince me that they sound good?
It's a losing battle, they are not to my taste and moreover (based on the little theory I know) I cannot understand why one should generate noise and then design a filter to shape it.
I know you like to design filters but I'm on the opposite side, I run DAC in NOS mode without any digital filter.
Or do you think one has to know perfectly the theory to understand how they sound?
But honestly I don't understand, do you want to convince me that they sound good?
It's a losing battle, they are not to my taste and moreover (based on the little theory I know) I cannot understand why one should generate noise and then design a filter to shape it.
I know you like to design filters but I'm on the opposite side, I run DAC in NOS mode without any digital filter.
In a PLL you have phase noise from the frequency reference, the VCO, the dividers, the phase detector and the loop filter. Usually the reference dominates and the VCO noise is much suppressed when you are well within the PLL loop bandwidth. If the close-in phase noise is degraded substantially by a PLL, its loop bandwidth is probably too small, that is, not so much greater than the frequency offsets of interest that the VCO noise is sufficiently suppressed.
Hello Andrea,
have you tried HQPlayer with x2, x4 or x8 upsampling with ext2 filter? It can change a bit your opinion about NOS' peremptory I suppose, if you didn't try yet.
My, absolutely personal, but rock solid experience is that on different occasions, different r2r dacs (MSB, Metrum, pcm1704, pcm63, pcm58) - having them put in oversampling mode through HQP - had turned their sound sensibly, explicitly better.
That is, feeding them instead of the original audio signal an oversampled, filtered and dithered (that is, agressively noise shaped) signal - the sound got much better.
So much for sound quality.
Also I have at home different dacs. Good old 1541, in an execution widely recognized.
It sounds good.
Until I turn on my ak4499 dac.
I am telling this to show that opinions and experiences can be strongly different. At this moment You could not pay me enough to turn back on my trails.
After 13 years with fantastic PCM63k.
Ciao, George
That is, feeding them instead of the original audio signal an oversampled, filtered and dithered (that is, agressively noise shaped) signal - the sound got much better.
So much for sound quality.
Also I have at home different dacs. Good old 1541, in an execution widely recognized.
It sounds good.
Until I turn on my ak4499 dac.
I am telling this to show that opinions and experiences can be strongly different. At this moment You could not pay me enough to turn back on my trails.
After 13 years with fantastic PCM63k.
Ciao, George
I can't tell anything about the complicated debate that is going on in this thread as i have not sufficient knowledge to do so; but i have a much higher experience about listening to various configurations of dacs and hifi systems.
If we have to stick to measurements and accordance to theoretical doctrine about electronics in DA conversion and clocking, maybe it's true that sigma-delta are some sort of microwave oven about RF noise; but in that viewpoint it's also true that giving so much importance to clock's close in phase noise is not totally justified.
Instead, if we have to rely on our ears, then forget any argument about sigma-delta dacs generating too much RF noise or multibit dacs having imprecise behaviour and we just have to judge upon how any technology honestly sounds.
I see many multibit-lover diy audiophile heavily customize their dacs, working on every little detail about power supplies, output stages, input signals, etc... to reach top of the class pleasant sound. Could it then be fair to compare those heavily optimized creations to sigma-delta medium (or low, like the Audiophonics Sabre dac Andrea tested) level commercial dacs? Clearly they are not meant for audiophiles that spend even (or maybe more than) 1000USD just for their clock setup...
Upon my experience, i find that if you compare a very well implemented multibit dac to a medium/low commercial sigma-delta dac (that is never optimized to a comparable level), the superiority of the former is simply clear, in terms of musicality, timbre and listening pleasure. BUT if we jump into the realm of heavily optimized sigma-delta diy builds, i sincerely have to admit that some very well implemented sigma-delta dacs based on recent chips from ESS or AKM are different beasts and they can give any oldish multibit dac i ever heard a serious competition even in listening pleasure, with very good timbre, naturalness and linearity - and holding quite a superiority in detail and threedimensional instruments separation. I wouldn't certainly use them as microwave ovens... No problem about "legendary" RF noise or similar, for me, if they sound so good!
I also have to admit that some very costly new R2R dacs are as good as sigma-delta dacs (or better) even in fields were the latters had a superiority...
So IMHO it's never a wise thing to stay still in a firm assumption and turn it in a strong and unquestionable prejudice; instead i find that this is easily made in audiophile forums, expecially (and this sounds very strange to me) by people who look very competent about audio technology.
About clocks, i can say this: in a 9038pro dac i have, i was originally using asynchronous mode, based on a onboard Crystek clock; then switched to a more complex setup using synchronous mode and Crystek-based MCK and reclocking. The improvement was huge and outstanding, even if the clocks i used were same brand, same model and obviously same phase noise. I also find the dac sounds different than using NDK SDA clocks in the same setup, even if phase noise is not in a different league. Not better sounding, not worse, simply different.
So i think that phase noise is probably an important factor, but there are other very important factors too, both in clocks themselves and (of course) in clocks interactions in your setup, power supply, reclocking, isolation and so on. Again, focusing too much on a single, strong assumption may be a limiting thing to do. Measurements are a powerful and very useful tool for designing refined systems, but obviously there's something more than good measurements to listening pleasure, or simply the factors involved are so many that it's very difficult to measure and predict them all.
We'll listen soon to these new SOTA clocks from Andrea: they seem very carefully designed and implemented, so i hope his assumption about phase noise is correct, time will tell. But i don't think it will be possible for us to perfectly predict what factors are more important until we can use our ears to verify, so please let him work in his path and let's hope he can prove to anybody that it's a good path
If we have to stick to measurements and accordance to theoretical doctrine about electronics in DA conversion and clocking, maybe it's true that sigma-delta are some sort of microwave oven about RF noise; but in that viewpoint it's also true that giving so much importance to clock's close in phase noise is not totally justified.
Instead, if we have to rely on our ears, then forget any argument about sigma-delta dacs generating too much RF noise or multibit dacs having imprecise behaviour and we just have to judge upon how any technology honestly sounds.
I see many multibit-lover diy audiophile heavily customize their dacs, working on every little detail about power supplies, output stages, input signals, etc... to reach top of the class pleasant sound. Could it then be fair to compare those heavily optimized creations to sigma-delta medium (or low, like the Audiophonics Sabre dac Andrea tested) level commercial dacs? Clearly they are not meant for audiophiles that spend even (or maybe more than) 1000USD just for their clock setup...
Upon my experience, i find that if you compare a very well implemented multibit dac to a medium/low commercial sigma-delta dac (that is never optimized to a comparable level), the superiority of the former is simply clear, in terms of musicality, timbre and listening pleasure. BUT if we jump into the realm of heavily optimized sigma-delta diy builds, i sincerely have to admit that some very well implemented sigma-delta dacs based on recent chips from ESS or AKM are different beasts and they can give any oldish multibit dac i ever heard a serious competition even in listening pleasure, with very good timbre, naturalness and linearity - and holding quite a superiority in detail and threedimensional instruments separation. I wouldn't certainly use them as microwave ovens... No problem about "legendary" RF noise or similar, for me, if they sound so good!
I also have to admit that some very costly new R2R dacs are as good as sigma-delta dacs (or better) even in fields were the latters had a superiority...
So IMHO it's never a wise thing to stay still in a firm assumption and turn it in a strong and unquestionable prejudice; instead i find that this is easily made in audiophile forums, expecially (and this sounds very strange to me) by people who look very competent about audio technology.
About clocks, i can say this: in a 9038pro dac i have, i was originally using asynchronous mode, based on a onboard Crystek clock; then switched to a more complex setup using synchronous mode and Crystek-based MCK and reclocking. The improvement was huge and outstanding, even if the clocks i used were same brand, same model and obviously same phase noise. I also find the dac sounds different than using NDK SDA clocks in the same setup, even if phase noise is not in a different league. Not better sounding, not worse, simply different.
So i think that phase noise is probably an important factor, but there are other very important factors too, both in clocks themselves and (of course) in clocks interactions in your setup, power supply, reclocking, isolation and so on. Again, focusing too much on a single, strong assumption may be a limiting thing to do. Measurements are a powerful and very useful tool for designing refined systems, but obviously there's something more than good measurements to listening pleasure, or simply the factors involved are so many that it's very difficult to measure and predict them all.
We'll listen soon to these new SOTA clocks from Andrea: they seem very carefully designed and implemented, so i hope his assumption about phase noise is correct, time will tell. But i don't think it will be possible for us to perfectly predict what factors are more important until we can use our ears to verify, so please let him work in his path and let's hope he can prove to anybody that it's a good path
Last edited:
Hi Andrea.
Don't get me wrong. The work you have been doing is absolutely fantastic.
I have several times expressed my thanks on behalf of diy community for your efforts and willingness to share.
With reference to close-in PN, your oscillators are very close to the SOTA units on the market today.
I have seen VERY few Wenzel & BVA' s that were slightly better, but they cost = a very nice BMW ��.
"Audio" oscillators generated a lot of hype some years ago but the hype has now died out (DAC chip improvements).
Producers (early 2000's) jumped on-board to make money off an ever collapsing Hi-End audio market & basically copied schematics off of ARRLA books and promised improved DAC performance (not).
Believe it or not, the first (person?) who mentioned close-in PN was Allan Wright RIP (I believe in the mid 90's).
He was a fierce promotor of very low close-in PN.
He always mentioned the importance of (sub 1 Hz) PN and that nobody seemed to care/understand this issue back then.
But i guess what I mentioned a cpl of years ago, and re-minded by Damian, is that lowering noise, any noise, either phase or amplitude, gets to a point of diminishing returns.
But I guess that's a little besides the point.
I for one really appreciate your hard work and your kindness to share the info.
I am not sure people realize how much time and effort it takes for what you are doing.
Keep up the good work! ������
Last edited:
I believe that debating around DAC technology is a little off topic in this thread.
Anyway we made a fundamental choice after years of listening to commercial and DIY devices: we are designing the audio system of our life without any compromis (in our opinion) and so we believe that the best way to achieve the best result is a multibit DAC.
But as I said we are designing a whole audio system, not only the master clock for the DAC.
So we are focusing on all the factors that in our opinion could do the difference to reach the best result.
Indeed we are designing SOTA oscillators, discrete multibit DACs, FIFO buffer and also source, ampli and speaker system.
In the meantime we also design a lite version of some devices, to get them more affordable to the audio community.
But our approach is always the same, the choise is done, therefore the way is established. Of course, the lite versions have inevitably some compromises.
Then there is no room for "balanced view" or better "balanced design".
We don't trust in versatile devices, when the choice is done we must follow the path we consider most suitable to reach the best result based on the initial choice.
Indeed, since we have chosen multibit DAC we think that the most crucial clock for this kind of DAC was the LRCK, so we decided to carry it to the DAC via copper dividing down directly from the master clock, and consequently the FIFO will be slaved to the LRCK.
On the other side we don't care much about the jitter of the bit clock and data because they don't affect the conversion, we only need bit perfect.
In the top version of our digital system bit clock and data come directly from the FPGA but they have to be strongly isolated from the LRCK, so we will use fiber optic cables to avoid any interference.
Finally, what I would say is that we cannot follow every expectation and design devices suitable for all configurations, simply because in our opinion it's not possible.
IMHO those who claims the above are promoting the philosopher's stone, but we well know that it doesn't exist.
You can argue if our fundamental choices are right or not, but by now the choice has been made and there is no chance to go back after years of development, and mostly in our opinion the delta-sigma DAC is not an option.
Anyway we made a fundamental choice after years of listening to commercial and DIY devices: we are designing the audio system of our life without any compromis (in our opinion) and so we believe that the best way to achieve the best result is a multibit DAC.
But as I said we are designing a whole audio system, not only the master clock for the DAC.
So we are focusing on all the factors that in our opinion could do the difference to reach the best result.
Indeed we are designing SOTA oscillators, discrete multibit DACs, FIFO buffer and also source, ampli and speaker system.
In the meantime we also design a lite version of some devices, to get them more affordable to the audio community.
But our approach is always the same, the choise is done, therefore the way is established. Of course, the lite versions have inevitably some compromises.
Then there is no room for "balanced view" or better "balanced design".
We don't trust in versatile devices, when the choice is done we must follow the path we consider most suitable to reach the best result based on the initial choice.
Indeed, since we have chosen multibit DAC we think that the most crucial clock for this kind of DAC was the LRCK, so we decided to carry it to the DAC via copper dividing down directly from the master clock, and consequently the FIFO will be slaved to the LRCK.
On the other side we don't care much about the jitter of the bit clock and data because they don't affect the conversion, we only need bit perfect.
In the top version of our digital system bit clock and data come directly from the FPGA but they have to be strongly isolated from the LRCK, so we will use fiber optic cables to avoid any interference.
Finally, what I would say is that we cannot follow every expectation and design devices suitable for all configurations, simply because in our opinion it's not possible.
IMHO those who claims the above are promoting the philosopher's stone, but we well know that it doesn't exist.
You can argue if our fundamental choices are right or not, but by now the choice has been made and there is no chance to go back after years of development, and mostly in our opinion the delta-sigma DAC is not an option.
I am quite sure that a high percentage of the people in your group buy are
planning to use their oscillators for DS DAC's. As such this discussion is very
relevant, however complex or difficult to understand.
They are part and parcel of the funding of this project regardless if you
only have a vested interest in your own MB DAC design.
It's interesting discussion and I think there may be ways to mitigate the
effects of cycle to cycle timing anomalies due to these sub harmonics. I'll go
into that tomorrow.
TCD
Terry,
I understand, but as I said we have not the time to develop whatever device, we have to focus on the approach we have chosen.
BTW, admitted and not granted that the sub-harmonics could affect the delta-sigma conversion, there are some options to solve the issue:
- use 22/24 MHz oscillators without any doubler (they are very good oscillators)
- add more filters to the output of the doubler
Simply we have no time to work more around the doubler, but everyone could do it by themselves.
What we will do is to analyze the spectrum of the doubler after the sine to square converter to understand if the sub-harmonics disappear or not.
I understand, but as I said we have not the time to develop whatever device, we have to focus on the approach we have chosen.
BTW, admitted and not granted that the sub-harmonics could affect the delta-sigma conversion, there are some options to solve the issue:
- use 22/24 MHz oscillators without any doubler (they are very good oscillators)
- add more filters to the output of the doubler
Simply we have no time to work more around the doubler, but everyone could do it by themselves.
What we will do is to analyze the spectrum of the doubler after the sine to square converter to understand if the sub-harmonics disappear or not.
As this clock is meant for digital audio, I'd say that discussions about the sensitivities of different types of audio DACs and ADCs to clock imperfections are on topic, while discussions about the relative merits of different types of audio DACs and ADCs are not. In any case, Andrea summarized the options you have when your DAC is sensitive to subharmonics very well in post #2877.
- Status
- Not open for further replies.